🚀 go-pugleaf

"Todd Knarr" <tknarr@silverglass.org> wrote in message
news:3FPd8.1643$2v6.233867@news1.west.cox.net...
> In cox.internet.discussion.athomeconversion
<_xOd8.3976$jC3.157556@news1.east.cox.net> Arthur Hagen <art@broomstick.com>
wrote:
> > If you're thinking of a NAT "router" as in the supermarket variety that
does
> > address translation and acts as a DHCP server, it solves the problem IF
and
> > only IF you don't need two IP addresses reachable from the net, or the
same
> > service running on two machine.  It will appear to be one machine to the
>
> Actually a garden-variety router _will_ solve thise problem, even if you
> have both machines with externally-reachable IP addresses. Just disable
> the NAT on it ( my Netgear RP114 can do that easily ).

With two machines on different subnets, will it route packets between them?
I believe all the sub-$500 routers out there aren't able to do so, but only
routes between the inside and outside, and acts like a dumb switch for the
LAN - not a true router.

And furthermore, how are the two machines going to get their IP from Cox?
If the router is in between, the only MAC that the cable modem sees is the
router's.  Or can you configure this router to have TWO MAC addresses on the
WAN side?
Or can your router be totally invisible and act like a smart switch?  That
won't help either, because then the two machines on the inside needs to know
to send the packets to each other and not through a gateway - in other
words, you need a static route or RIP information, which isn't provided, and
you're back to zero.

Have you *tried* it?  :-)

> > anything about this situation.  If Cox had provided RIP information like
> > @home did, it could.  However, I have received email from Cox stating
that
> > they will NOT provide RIP information, so this won't work either.
>
> I do it without RIP. All you need to do is enter the static routes into
> the router.

Which is kind of hard to do with dynamic and changing IPs, wouldn't you say?

> In fact I do this even when the ISP provides RIP, because
> a) the routing isn't that complex, I have exactly 1 internal network and
> everything else is on the WAN ( translation: 1 normal route for the
> internal subnet, 1 default route ) and b) RIP provides no strong security
> or protection against forged updates, so accepting RIP allows someone
> else on Cox's network to update your routing tables to run your LAN's
> traffic through their machine.

Ah, but here's the caveat - when you get TWO (or more) IPs from Cox, they
will NOT be on the same subnet.  So you have several subnets on the inside,
which is the whole problem.  Machine A believes that machine B is only
reachable through the default gateway, and sends the packets to the gateway.
Which in your case is on the *outside*.  Machine A needs to have a static
route to machine B, which isn't easy to automate because of the dynamic
DHCP.

As for RIP information, the way it works is that the client sends out a RIP
*request*, which is answered by the routers on the local net.  This request
is NOT forwarded.  The router in turn sends its *own* requests to the next
step.  You can't push your false RIP info to someone else, because it's a
pull operation.  And even if a RIP request somehow got forwarded to the
outside (i.e. to other Cox users on your subnet), they would have to be able
to send a spoofed reply *before* the router.  With the latency, this simply
won't happen (unless it's a router that's *really* fast at routing, and
*really* slow at answering RIP requests, and for some unfathomable reason
routes RIP requests meant for itself).
With RIP2 there's even more security, with sequence numbers and verification
of RIP announcements.

Back to the real problem:  If anyone has successfully managed to set up
direct TCP/IP between two or more Cox assigned IP addresses (i.e. *not*
using NAT) on different subnets behind the same cable modem, and without
manual static route configuration every time DHCP info changes, I'd like to
hear about it.  Short of adding a second NIC to each box, or buying a CISCO
Catalyst switch, that is.

Regards,
--
*Art